Automated Learning Pipeline for Antimicrobial Resistance
(Orange parts are not available in lite version)
Single-Reference AMR is installable from conda using mamba:
To install it into the new environment:
mamba create -n alpar -c conda-forge -c kalininalab -c bioconda -c etetoolkit alpar
conda activate alpar
pip install panacotaOr to install it into the already existing environment:
mamba install -c conda-forge -c kalininalab -c bioconda -c etetoolkit alpar
pip install panacotaALPAR Lite is installable from conda using mamba:
Lite version of the tool, missing; DataSAIL (Split Against Information Leakage), Pyseer (GWAS), PanACoTA (Alignment-Based Phylogenetic Tree Creation), MashTree (Alignment-Free Phylogenetic Tree Creation), PRPS (Phylogeny Related Parallelism Score), AutoSklearn (Machine Learning Parameter Optimization)
For more information : ALPAR-Lite
Example files can be downloaded from:
From genomic files, creates binary mutation and phenotype tables, applies thresholds, creates phylogenetic tree, conducts GWAS analysis, calculates PRPS score and trains machine learning models with conducting feature importance analysis and splitting data aginst information leakage with DataSAIL against all the given antibiotics.
-
Input,
-i: Path of folder that have structure: input_folder -> antibiotic -> [Resistant, Susceptible]📦input_folder ┣ 📂antibiotic1 ┃ ┣ 📂Resistant ┃ ┃ ┣ 📜fasta1.fna ┃ ┃ ┗ 📜fasta2.fna ┃ ┃ ┗ ... ┃ ┗ 📂Susceptible ┃ ┃ ┣ 📜fasta3.fna ┃ ┃ ┗ 📜fasta4.fna ┃ ┃ ┗ ... ┗ 📂antibiotic2 ┃ ┣ 📂Resistant ┃ ┃ ┣ 📜fasta2.fna ┃ ┃ ┗ 📜fasta5.fna ┃ ┃ ┗ ... ┃ ┗ 📂Susceptible ┃ ┃ ┣ 📜fasta2.fna ┃ ┃ ┗ 📜fasta3.fna ┃ ┃ ┗ ... ┗ 📂...
-
Output,
-o: Output folder path, where the output will be stored. If path exist,--overwriteoption can be used to overwrite existing output. -
Reference,
--reference: Reference file path, accepted file formats are:.gbk .gbff -
Custom database (Optional),
--custom_database: Fasta file path for protein database creation, can be downloaded from UniProt accepted file formats are:.fasta
Basic usage:
alpar automatix -i example/example_files/ -o example/example_output/ --reference example/reference.gbffFrom genomic files, creates binary mutation and phenotype tables
-
Input,
-i: Path of file that contains path of genomic fasta files per line or path of folder that have structure: input_folder -> antibiotic -> [Resistant, Susceptible]📦input_folder ┣ 📂antibiotic1 ┃ ┣ 📂Resistant ┃ ┃ ┣ 📜fasta1.fna ┃ ┃ ┗ 📜fasta2.fna ┃ ┃ ┗ ... ┃ ┗ 📂Susceptible ┃ ┃ ┣ 📜fasta3.fna ┃ ┃ ┗ 📜fasta4.fna ┃ ┃ ┗ ... ┗ 📂antibiotic2 ┃ ┣ 📂Resistant ┃ ┃ ┣ 📜fasta2.fna ┃ ┃ ┗ 📜fasta5.fna ┃ ┃ ┗ ... ┃ ┗ 📂Susceptible ┃ ┃ ┣ 📜fasta2.fna ┃ ┃ ┗ 📜fasta3.fna ┃ ┃ ┗ ... ┗ 📂...
-
Output,
-o: Output folder path, where the output will be stored. If path exist,--overwriteoption can be used to overwrite existing output. -
Reference,
--reference: Reference file path, accepted file formats are:.gbk .gbff -
Custom database (Optional),
--custom_database: Fasta file path for protein database creation, can be downloaded from UniProt accepted file formats are:.fasta -
Creation of phenotype table (Optional):
--create_phenotype_from_foldershould be used- Genomes_folder_path should have a structure: input_folder -> antibiotic -> [Resistant, Susceptible] -> genomic fasta files
📦input_folder ┣ 📂antibiotic1 ┃ ┣ 📂Resistant ┃ ┃ ┣ 📜fasta1.fna ┃ ┃ ┗ 📜fasta2.fna ┃ ┃ ┗ ... ┃ ┗ 📂Susceptible ┃ ┃ ┣ 📜fasta3.fna ┃ ┃ ┗ 📜fasta4.fna ┃ ┃ ┗ ... ┗ 📂antibiotic2 ┃ ┣ 📂Resistant ┃ ┃ ┣ 📜fasta2.fna ┃ ┃ ┗ 📜fasta5.fna ┃ ┃ ┗ ... ┃ ┗ 📂Susceptible ┃ ┃ ┣ 📜fasta2.fna ┃ ┃ ┗ 📜fasta3.fna ┃ ┃ ┗ ... ┗ 📂...
Basic usage:
alpar create_binary_tables -i example/example_files/ -o example/example_output/ --reference example/reference.gbffApplies threshold to binary mutation table, and drops columns that has less than threshold percentage, useful to reduce sequencing errors in the data.
-
Input,
-i: Binary mutation table path -
Output,
-o: Output folder path, where the output will be stored. If path exist,--overwriteoption can be used to overwrite existing output. -
Threshold percentage,
--threshold_percentage: Threshold percentage value to be used to drop columns. If column sum is less than this value, columns will be deleted from table
Basic usage:
alpar binary_tables_threshold -i example/example_output/binary_mutation_table.tsv -o example/example_output/ Runs Phylogeny pipeline to create phylogenetic tree. (Alignment free)
-
Input,
-i: Text file that contains path of each strain per line. It can be found in create_binary_tables output path asstrains.txt -
Output,
-o: Output folder path, where the output will be stored. If path exist,--overwriteoption can be used to overwrite existing output. -
Random names dictionary path,
--random_names_dict: Random names text file path. If not provided, strain's original names will be used for phylogenetic tree
Basic usage:
alpar phylogenetic_tree -i example/example_output/strains.txt -o example/example_output/ --random_names_dict example/example_output/random_names.txt Runs PanACoTA pipeline to create phylogenetic tree. (Alignment based)
-
Input,
-i: Text file that contains path of each strain per line. It can be found in create_binary_tables output path asstrains.txt -
Output,
-o: Output folder path, where the output will be stored. If path exist,--overwriteoption can be used to overwrite existing output. -
Random names dictionary path,
--random_names_dict: Random names text file path. If not provided, strain's original names will be used for phylogenetic tree
Basic usage:
alpar panacota -i example/example_output/strains.txt -o example/example_output/Runs GWAS analysis to detect important mutations in the data.
-
Input,
-i: Binary mutation table path that is created via create_binary_tables command, can be found in create_binary_tables output path asbinary_mutation_table_with_gene_presence_absence.tsvorbinary_mutation_table.tsvor if threshold applied, can be found in binary_table_threshold output path asbinary_mutation_table_threshold_*_percent.tsv -
Phenotype,
-p: Binary phenotype table path, can be found in create_binary_tables output path asphenotype_table.tsvif--create_phenotype_from_folderis used. Can also created manually and used. -
Tree,
-t: Phylogenetic tree path, can be found in panacota output path asphylogenetic_tree.newickor phylogeny output path asphylogenetic_tree.tree -
Output,
-o: Output folder path, where the output will be stored. If path exist,--overwriteoption can be used to overwrite existing output.
Basic usage:
alpar gwas -i example/example_output/binary_mutation_table_with_gene_presence_absence.tsv -p example/example_output/phenotype_table.tsv -t example/example_output/phylogeny/phylogenetic_tree.tree -o example_output/Runs PRPS (Phylogeny-Related Parallelism Score) to detect the mutations are more likely associated with phylogeny rather than antimicrobial resistance.
-
Input,
-i: Binary mutation table path that is created via create_binary_tables command, can be found in create_binary_tables output path asbinary_mutation_table.tsvor if threshold applied, can be found in binary_table_threshold output path asbinary_mutation_table_threshold_*_percent.tsv -
Tree,
-t: Phylogenetic tree path, can be found in panacota output path asphylogenetic_tree.newickor phylogeny output path asphylogenetic_tree.tree -
Output,
-o: Output folder path, where the output will be stored. If path exist,--overwriteoption can be used to overwrite existing output.
Basic usage:
alpar prps -i example/example_output/binary_mutation_table.tsv -t example/example_output/phylogeny/phylogenetic_tree.tree -o example_output/Trains machine learning models with classification algorithms on the data and optimizes.
Available Classification algorithms: Random Forest, Support Vector Machine and Gradient Boosting
-
Input,
-i: Binary mutation table path that is created via create_binary_tables command, can be found in create_binary_tables output path asbinary_mutation_table_with_gene_presence_absence.tsvorbinary_mutation_table.tsv -
Phenotype,
-p: Binary phenotype table path, can be found in create_binary_tables output path asphenotype_table.tsvif--create_phenotype_from_folderis used. Can also created manually and used. -
Output,
-o: Output folder path, where the output will be stored. If path exist,--overwriteoption can be used to overwrite existing output. -
Antibiotic,
-a: Antibiotic name that model will be trained. Should match the name with column that represents phenotype in binary phenotype table -
Optional arguments:
- Machine learning algorithm,
--ml_algorithm: Classification algorithm to be used, available selections: [rf, svm, gb] - Resampling strategy,
--resampling_strategy: Resampling strategy to be used, available selections: [holdout, cv] - Parameter optimization,
--parameter_optimization: Parameter optimization for model with autosklearn (https://automl.github.io/auto-sklearn/master/index.html) - Save model,
--save_model: Save model - Feature importance analysis,
--feature_importance_analysis: Analyze important features in the model with permutation importance - Datasail,
--sail: Splits data into training and test sets against information leakage to train better models. Requires text file that contains path of each strain per line. It can be found in create_binary_tables output path asstrains.txt
More optional arguments can be found in help page:
python alpar/sr_amr.py ml -h
- Machine learning algorithm,
Basic usage:
alpar ml -i example/example_output/binary_mutation_table.tsv -p example/example_output/phenotype_table.tsv -o example_output/ -a amikacin
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent